Alex Beutel's Blog

Time Series Twitter

June 12th, 2012 · 1 Comment

Recently my friend Josh Hammer has been sending a TON of tweets.  He is a staunch Republican and Mitt Romney supporter.  If you want to read every news story through the GOP lens, subscribe to his Twitter feed.  He is effectively a GOP Twitter bot.

In the past couple of weeks, his tweeting began to get out of control, and I wondered: at this rate, how many tweets will he send by Election Day?  Taking what I have learned from following his Twitter feed and reading conservative news, I took a small amount of data and irresponsibly extrapolated it.  Thus, Time Series Twitter was born.

Time Series Twitter is a small, simple web app – it simply loads the last 1000 tweets from a given user, graphs the time series plots for the tweets (by day), and does a linear regression on the tweets.  With the linear regression you can over extrapolate and find the expected number of tweets a given user will send some time in the future.

For Hammer, this is currently 57 Tweets per day by the Election, significantly more than other news sources: @FoxNews is expected to send 6.3 tweets/day, @ESPN is expected to send 10.9 tweets/day, and @MSNBC to send -36.2 tweets/day. (All of these measurements are at time of writing – they obviously change a lot.)

This web app is nothing novel (or accurate), but it has been fun to watch friends’ Twitter activity over time.  Give it a try.

Tags: Javascript · Miscellanious · Twitter

1 response so far ↓

  • 1 Niru // Jun 13, 2012 at 12:20 am

    looks awesome, nice work

Leave a Comment